Skip to main content
SearchLogin or Signup

Automating Abandonment

Published onNov 24, 2020
Automating Abandonment

Efficiency is a specter often beckoned in appeals to automation. The computerization of the welfare state, which began in the 1970s, was celebrated as an attempt to make the system more efficient. A lean, digitally-mediated bureaucracy offers the seductive promise that the state will save public funds, serve welfare recipients better by cutting out subjective decision-making and corruption, and simplify processes as diverse as submitting applications and filing claims.

The political demand from which the welfare state grew was guided by the principle of incrementally overcoming unnecessary suffering. While such a principle might once have been the guiding light of a certain kind of reformist, gradualist labor movement, the precarity of life is unavoidable. As we age, the odds against survival grow. There will be suffering, there will be precarity – to imagine otherwise is rather too romantic and utopian.

In much of the enriched world and even beyond it, reformist labor movements struggled to embed this kind of ethical socialist principle in forms of administered and institutional care that collectively came to be called the welfare state. Such investments in public care led to the construction of national health systems, but also public assistance for aged care, disability care, and so on. It wasn’t restricted to the elderly either. In some countries, it extended across all life stages, from birth through education, work, retirement, and death.

The weakness of the American labor movement, rife with racism, meant that this program was not implemented all that thoroughly in that most enriched of rich worlds. Still, mass-scale social engineering projects such as the New Deal and the Great Society sought to alleviate poverty by supplementing the wage with entitlements, and even providing the wage itself in the case of unemployment, retirement, or disability.

That said, Precarity Lab is not particularly nostalgic for the post-war social welfare state. Like most institutions, those of the social welfare state embodied mixed and even antithetical agendas. Social welfare functioned mainly to reproduce labor so that labor might reproduce capital. It was riddled with racial exclusions. It insisted on normative models of gender and sexuality. It enshrined the patriarchal family model as a norm. And – no surprise – welfare states were also laboratories. They were sites of experiment on the physical and mental attributes of bodies. Welfare disciplined potential recipients by making eligibility conditional on adherence to respectability, a heteronormative family structure, and work requirements. Administrators and experts served in gatekeeping functions, obliging precarious people in particular to ventriloquize the comportment and language that would open the gate to housing assistance, a scholarship, or access to medical gender transition.

Race, gender, ability, and sexuality have been used to distinguish between the “eligible” and “ineligible.” The safety net always had holes big enough for certain categories of people, already living precarious lives, to fall through.

Two generations of austerity governance eroded even these compromised forms of the social welfare state in those parts of the enriched world that might somehow have afforded them. No longer is the incremental elimination of unnecessary suffering the guiding light for all to see. But there may be more going on here than just the neoliberal shift to market-based solutions and austerity rationing.

These days, one might take an even less “charitable” view of the purpose and function of the social welfare state, particularly given recent experiments with technology designed to update and automate the management of public assistance. One could even ask if what is left of the social welfare state no longer even has the mission of eliminating unnecessary suffering, but rather of using digital surveillance techniques to exacerbate suffering as a means of control and rent extraction.

Automating the Sluggish Inefficiencies of Bureaucracy

Algorithmic decision-making replaces the human, and sometimes more humane, discretion of state bureaucrats such as caseworkers and claims processors.1 Humans are bad at calculating data compared to machines, but they can be negotiated with – they can have compassion. This is not to say that they always do – their decisions may not be always less racist or sexist than algorithms – but they can.

Yet these fantasies of the technological “quick fix” to the expensive sluggishness of government decision-making obscures a fundamental question: efficiency for whom, and at whose expense? The automation of public benefit administration acts as a covert austerity accelerant that hollows out social programs, while allowing a shell of the various programs comprising the welfare state to persist in name alone.

A US Department of Agriculture fraud detection algorithm recently determined that a New York City grocer was processing food assistance payments after already providing items from his store to customers on credit.2 The grocer would allow community members to get groceries on credit when they had already exhausted their benefits, and charge them for what they needed to scrape by until their benefits were replenished for the month. The grocer was not lauded for providing community members with foods necessary to survive, and instead was barred from participating in the SNAP food assistance program altogether. Not only did this decimate the grocer’s income, but also broadly affected the lives of low-income members of his community. Here, the undergig that enables benefits to be distributed without record and surveillance is punished for providing care, a necessary supplement when social welfare is utterly broken.

By automating bureaucracy, “benefits” can act as techniques of extraction in neoliberal sheep’s clothing, covering over the sly dismantling of the welfare state under the guise of efficient technocratic management. All too conveniently the digital technologies designed to render bureaucratic labor obsolete reproduce existing structural inequalities at best, if they do not kick people off of public assistance programs altogether.

The Subject of Public Assistance from Eligible Recipient to Beneficiary

The disciplining function of determining who is eligible for benefits also continues to persist under conditions of automated efficiency. Our allure with disrupting bureaucracy with tech-driven efficiency distracts us from the impact of a system that decides whether or not real people are “eligible” or “ineligible” for receiving public support on top of layers of abstraction that are ultimately reducible to a human-free logic to render binary profiles “ineligible.”

Machine learning algorithms might give the binary a valence of probability, but ultimately a threshold of resemblance determines whether or not one is included within the set of the eligible. Because artificial intelligence is ultimately just computationally juiced-up statistical analysis, one is at first never entirely eligible or ineligible. Instead, data about individual behavior is determined to correspond more or less to predetermined statistical models of eligibility and ineligibility. If one’s profile corresponds to the eligible model beyond a certain degree of probability exceeding a minimum threshold, say, 95 percent, they will be counted as eligible.

Already precarious “beneficiaries” are often kicked out of welfare programs and disallowed from negotiating with the state offices that claim to serve them. Transferring authority to make decisions about public assistance eligibility from people to machines makes negotiating with those accountable for making these determinations seem impossible. The person abandoned by the automated welfare state confronts a new object, the clunky state interface, and is consequently no longer a subject precisely because they have been barred from predication, from conversation and negotiation – most often, not for the first time.

For example, if a Medicaid patient makes it through the eligibility elimination round of algorithmic determination, she faces the patient portal. Patient portals and online forms are maddeningly and terrifyingly exacting; they require specific types of browsers or apps that don’t work well on specific types of phones; tiny keyboards don’t lend themselves to typing strings of numbers; bodies already burdened with urgency – pain, debility, collections – aren’t easily capable of executing this difficult dance, which must be done perfectly every time. The tiny link at the bottom of the page that invites the sick and frustrated patient to get help solving problems using the site leads to an email address or phone number.

Appeals to the grace of overburdened state employees often turn out to be another inconvenient email in a bloated inbox that is slowly addressed; calling these phone numbers places one in a seemingly endless queue, waiting anxiously, feeling as if the welfare apparatus will never respond to calls for help.

As such, private health insurance is not protection. Patients may abruptly be denied coverage of a specific medication or doctor’s visit on the grounds their paperwork was not filled out perfectly, or a doctor that was listed as “in-network” is actually “out-of-network.” And the denial of service – a distributed denial-of-service attack, if you will – affects the would-be service providers and the would-be served; in some cases, human doctors and nurses who have trained for years to protect human life and address suffering fight to get to patients just as hard as we fight to get to them for care.

Accomplished by using opaque algorithms to flag people as “non-compliant,” the shrinking of the pool of eligibility by welfare offices moves us towards welfare programs without welfare recipients. The welfare state hasn’t gone away, exactly; it has merely been hollowed out by the implementation of streamlined bureaucratic obstructions.

Loan Forgiveness

A reminder: precarity is associated with the act of begging for what is necessary for one’s survival. The phrase “loan forgiveness” has precarity built into it, for the state is framed as the magnanimous absolver of the borrower’s debt. Yet in the case of public-service loan forgiveness, it is the indebted borrower who promises their labor to the state in the hope of eventually being free from debt.

Consider, for example, the promise of education proffered by student loans. Created in 2007, the Public Service Loan Forgiveness program in the United States enabled full-time public-sector and non-profit employees to have their student loans forgiven after ten years of payments. To save money, the US Department of Education contracted FedLoan to manage the PSLF program and repayment plans. All those enrolled in PSLF repayment plans were automatically switched to the private student debt management corporation, which is now being sued for misleading and exploiting borrowers; 99 percent of borrowers who applied for loan forgiveness under the PSLF program were rejected, often because they were tracked into repayment plans that rendered them ineligible for forgiveness, or they were rejected on technical grounds.3

The selling-out of educational loan forgiveness, intended to reward the use of education for public service, to a for-profit company, illustrates the irony of the turn to technocratic governance. Automation of social welfare program benefit-analysis has been sold to the public sector as an easier, time-saving alternative to bloated bureaucracies. In reality, automation multiplied obstacles and shrank “eligibility” on technical grounds.4 In addition, automation demanded a kind of relentless and time-consuming labor of countering the effects of being spit out as ineligible, labor for which many public-service-sector workers have no time.

Lured by the false promise of uplift through education, the worker laboring to pay off an often-predatory loan required as a condition of getting the very credentials required for employment by the state is led to take debt by unclear information. Workers are being screwed over by the far-from-benevolent state, both on the axis of the wage and the axis of debt. The spirals of precarity tighten with the turn of the screw.


Medicaid, part of that “third-rail” of American politics that is deemed untouchable and yet covered in cost-cutters’ fingerprints, provides a similar example. As part of the Affordable Care Act (ACA) in 2010, the federal government gave states the option of expanding Medicaid to legal permanent residents under 65 years old, whose incomes remained less than 133 percent of the federal poverty level (FPL). However, the ACA also allowed states to experiment with their own processes of providing public health care to their residents. In some cases, this might involve imposing cost-sharing requirements on those who already have trouble making ends meet, and in other states requiring Medicaid recipients to prove that they are spending several hours per month either working or acquiring training that would allow them to more easily find a job.

The State of Michigan, for example, refers to those who are on Medicaid as “beneficiaries” rather than “recipients” in all of its official communications. We would not refer to lab subjects or experimental animals as “beneficiaries” because they are providing data to enrich institutions. Medicaid has many lab-like characteristics, including automation and experimentation on vulnerable subjects.

States with Medicaid waivers experiment with methods to create a welfare state without welfare recipients. This is not governance by algorithm, but governance under cover of algorithm. As described earlier, Medicaid is not a benefit, but a process for refusing benefits. While Medicaid work requirements appear to target all Medicaid recipients, they primarily single out and target the 6 percent of Medicaid recipients who are currently out of institutionally recognized forms of work; 62 percent of recipients already work full- or part-time, and the 32 percent that are unable to work are in this category due to “qualified” disabilities or obligations related to school or caregiving. Unemployed people who live in rural counties or former-industrial cities that simply lack the jobs necessary to sustain their populations are likely to lose insurance. Moreover, those who do find employment, and work 40 hours a week at the federal minimum wage, will already make 125 percent of the FPL. A small raise or scheduled overtime thus puts recipients of public benefits at risk of losing their insurance.

Medicaid work requirements, along with other related technical “solutions” to the supposed crisis of the welfare state, co-constitute the change in the conceptualization of “welfare” – from a social right to an earned, and in many cases unearnable, reward. Technocratic governance has been sold to us as an easier, time-saving alternative to bloated bureaucracies. Instead, it has led to the multiplication of obstacles and the shrinking of “eligibility” on technical grounds. Anyone navigating the health care system, the welfare system, or student loan services today is intimately familiar with the endless quest to remain eligible for these programs by meeting requirements, acquiring and submitting accurate personal information, or correcting errors in their profiles.

The automation of the welfare state seeks to remove government workers and beneficiaries from the social safety net system. Automation embeds in the system a recourse to passive-voice deniability. “Your health care visit was not approved.” “Your public loan was not forgiven.” Automated decisions and automated statements pile on each other as if emanating from a faceless, ethereal substance. Within such technopolitics, higher education and health care are not rights: they are unexecutable code.

Who benefits when caseworkers are fired and no one is getting health care and/or is unable to pay off loans? Insurance companies, people in high tax brackets, private loan companies, and above all the massive digital services companies such as Accenture and SAP that program and maintain these systems of digital deniability.

Keeping People Alive in Order to Extract from Them

Exclusion-by-automation is only one aspect of the widening holes in the safety net. An ever more parasitic state and a predatory credit system is obliged to keep people alive in order to retain them as host bodies for extraction. While acknowledging the violence of exclusion from systems of basic life-sustenance, we also note the way that, at least in a US context, the privatization or private “partnership”-ization of state services renders “beneficiaries” as raw materials for profit.

The “innovation” of today’s mode of governance is not just to contract with private corporations to administer state programs, but actually to create pathways for the contracted theft of public resources. Private “administration” or “partnership” models of governance emerge out of a belief that state actors are ill-equipped to manage complex human systems such as health care, telecommunications, or mining – a belief, it should be noted, to which we are sympathetic, but from a different ideological position! – and that taxation of the owning class is immoral (a belief for which we have no sympathy).

A brief list of such “public” goods that have been transformed into captive markets for corporations through state consent include student loans, the transformation of public housing in Section 8 vouchers, state funding for charter schools, the expansion of the provision of “managed” mental and physical health services through private health insurance companies, and the fusion of shareholder-driven extraction and transmission investment and customer-facing utilities provision. As such, state services such as Medicaid or notionally state-regulated industries such as electrical utilities become clients/captive buyers for corporate actors.

In the absence of pressure from powerful social movements from below, the social welfare state has become a parasitic public built on the backs of people of color and women. But in the age of managed care, the privatization of education, private student loan collection, and private welfare administration, the public no longer exists. Through contracted theft, the managed state turns the “social safety net” itself into a site of extraction. It is a mechanism for transferring wealth to the private sector and funding expenditure in a fiscal climate where government bodies can no longer generate sufficient revenue through taxation.

States fail to generate revenue through taxation; increasingly, they actually transfer revenue smoothly into the private sector. Tax refunds are a lifeline for the 78 percent of Americans who live from paycheck to paycheck. Yet, in 2019, the federal government seized $3.3 billion from student borrowers’ federal tax refund for unpaid student debt, up from $2.3 billion in 2016 – a $1-billion increase in a short period of time. This reveals the degree to which parasitic governance is accelerating. While “permanently disabled” borrowers are eligible for student debt forgiveness, private debt collectors such as Navient create obstructions that make it difficult for indebted disabled people to discharge their student debt.

Corporate actors say they are acting in partnership with states. They say they are losing potential earnings by agreeing to offer services at fixed costs and foregoing the potentially lucrative vicissitudes of the open market. What actually happens is that these managed/administration arrangements create captive buyers – in the case of prisons, literally captive – which insulates private companies from risk and guarantees them not so much profits as rents.

Regardless of what they say to the public or to the state, these same companies tell shareholders that they can expect a stable return on investment, and insulation against loss of market share or collapse. This is a mode of governance whose mission is the welfare of the corporation. It is the corporation and its institutional investors who are shielded from the particular form of precarity that is the “free” market. Even if services are offered at cost – which, for the record, they almost never are, as private actors are generally permitted to establish “fair” margins of return in exchange for their “labor” administering state services – they still bolster the corporation’s earnings. States participate in this transfer of wealth by charging consumers, or by funneling consumers into particular fixed marketplaces, while the corporation earns the benefits.

The human being who seeks medical care, who wants to call a family member while incarcerated, who wants natural gas to heat their home, who takes out a student loan to pay for school – that person is the raw material from which corporations extract secured rents, all while dressed in the clothing of the state providing a safety net. The automation of the welfare state uses the ruse of “efficiency” to dismantle the so-called social safety net, producing precarity while claiming to provide security.


No comments here